A 3D bar or pie chart's perspective systematically makes some values look bigger than they are

What it is

Adding a third dimension to a chart (3D bars, 3D pies) is not "one more decoration" — it systematically injects error into magnitude reading: perspective projection makes near bars look taller than equally tall far bars, and 3D pie sectors are stretched or compressed by tilt, so equal values no longer read as equal. The third dimension adds no information — the data is still two-dimensional — only distortion.

Why it happens

The mechanism lives in projection geometry: the screen is two-dimensional, so 3D shapes must be projected onto it, and perspective projection scales by distance — the same value drawn deeper comes out smaller while readers apply flat rules of "equal height / equal area," with error growing linearly with depth difference. 3D pies pile on elliptical compression: tilting the disc shortens the vertical radius, stretching the fore and aft sectors unevenly, making an already high-error angle channel worse. The third dimension also hides information: with bars arranged front to back, the front row naturally occludes the back by an amount set by camera position, not data — the "comparison" readers perform is contaminated with the arbitrariness of the viewpoint.

Where it stops holding

3D does not fail everywhere: when the data itself is three-dimensional (spatial coordinates, molecular structure, terrain), 3D is the native form, not decoration; rotatable interactive 3D lets readers cancel viewpoint error at the price of operating burden. Static 2D data rendered in 3D enjoys no such exemption — every "extra dimension" benefit lives in the perception layer while every cost lands in accuracy. The criterion is "does the third dimension carry data": if not, it is decoration and falls to the next card's rules.

Applying it

• Default 2D data to a ban on 3D rendering: bars, pies, and ribbons ship flat; a 3D request requires data that is genuinely three-dimensional.
• When 3D is unavoidable (spatial data), provide orthographic projection or a fixed standard viewpoint and declare the viewpoint in the caption.
• Verification: place the 3D chart beside its 2D counterpart and have readers compare arbitrary pairs; markedly higher error in the 3D version is measured distortion — revert to 2D.

Related

• Same group: U5.07.2 Decorative elements crowd out reading attention · U5.07.3 Visual appeal must not be bought with accuracy
• Nearby: U3.04.1 Dual-axis relative position is a human choice · U2.06.1 Bubbles carry a third variable with the lowest reading accuracy of common channels
• Search terms: 3D chart distortion · perspective projection · 2D vs 3D accuracy